CPRD Back Pain Analysis Technical Document V4.1

CPRD Back Pain Analysis Technical Document v4.1

Stroke prevalence model for small populations:

Technical Document produced for Public Health England

Ben Hollis,Roger Newson, Bowen Su, Ioanna Tzoulaki, Pankaj Sharma, Azeem Majeed Michael Soljak

Department Primary Care & Public Health

School of Public Health

1

CHD/stroke/PAD prevalence modelling technical document

1Contents

1Executive Summary

2Background

2.1CVD Risk Factors

2.1.1Age

2.1.2Gender

2.1.3Ethnicity

2.1.4Hypertension

2.1.5Smoking

2.1.6Diabetes

2.1.7Obesity

2.1.8Dyslipidaemia

2.1.9Deprivation and socioeconomic status

2.1.10Physical Activity

2.1.11Inflammatory Markers

2.1.12Chronic Kidney Disease

2.1Stroke prevalence from the literature

3Methods

3.1“Data discovery” from UK survey data sources

3.1.1Whitehall II

3.1.2Airwave Health Monitoring Study

3.1.3Health Survey for England 2013

3.1.4English Longitudinal Study of Ageing

3.2Definition of Outcomes

3.3Risk factor variables

3.3.1Age

3.3.2Sex

3.3.3Ethnicity

3.3.4Hypertension

3.3.5Dyslipidaemia

3.3.6Family History (CHD)

3.3.7Smoking

3.3.8Obesity

3.3.9Diabetes

3.3.10Chronic Kidney Disease

3.3.11“Hypercoagulable state”/on anti-coagulant

3.3.12Inflammatory Markers

3.3.13Socioeconomic status/deprivation

3.3.14Physical Activity

3.4Statistical Analyses

3.4.1Whitehall II missing data

3.4.2Whitehall II descriptive analyses

3.4.3Whitehall II regression modelling

3.4.4Interactions

3.4.5Internal validation

3.5Local prevalence estimates

3.5.1Method 1: bootstrapping procedure to produce repeated samples

3.5.2Method 2: Logistic regression and inverse probability weights

3.6Validation of local estimates

3.6.1Internal validation

3.6.2External validation

4Results

4.1Stroke prevalence from Whitehall II Data

4.1.1Missing data

4.1.2Baseline characteristics of Whitehall II Respondents

4.2Whitehall II CVD definitions, incidence & prevalence

4.2.1Whitehall II prevalence

4.3Stroke regression modelling

4.3.1Univariate logistic analysis

4.3.2Multivariate logistic analysis for stroke

4.4Internal validation

4.4.1ROC curves

4.4.2Probability and sensitivity/specificity analysis

4.5Local estimates

4.5.1Internal validation

4.5.2External validation of practice estimates against QOF prevalence

5References

6Appendix: additional information

1

Stroke prevalence model 2016 Technical Document v1.1

Stroke prevalence model Technical Document

1Executive Summary

2Background

The Department of Primary Care & Public Health (PCPH) in the School of Public Health (SPH) at Imperial College London (ICL) has tendered successfully to Public Health England (PHE) to develop small population prevalence models for several chronic diseases. PHE has requested a single cardiovascular disease (CVD) prevalence model. CVD is a term used to define disorders which affect the heart and blood vessels. These may be broadly subcategorised into two classes: 1) atherosclerotic CVDs including coronary heart disease (CHD), cerebrovascular disease more commonly called stroke, and transient ischaemic attacks (TIAs), and peripheral arterial disease (PAD); and 2) CVDs of other aetiology including congenital and rheumatic heart disease and cardiac arrhythmias. We considered that it would be feasible, and more useful at the local level, to develop three separate prevalence models for CHD, stroke and TIA, and PAD, nevertheless using the same data source and exactly the same methods. There is a large but not complete overlap in risk factors for the three diseases, but the prevalence of each is distinct. The prevalence of each disease in each local population can be summed to provide an overall prevalence of CVD. (For the present models, we focus on the atherosclerotic CVDs (CHD, CD/stroke, and PAD) as these represent the vast majority of the CVD disease burden in the UK.) This Technical Document considers all CVD risk factors, but then describes the derivation of the model for CHD.

Collectively, CVD is the leading cause of mortality worldwide, accounting for 17.5 million deaths in 2012.[1] In the UK, CVD remains among the leading causes of death and presents a tremendous financial burden, with the National Health Service (NHS) having spent £6.8 billion on CVD treatment in 2012/13.[2] The prominent risk factors for CVD are well established. While many of these cannot be altered (e.g. age, gender, family history) there are numerous examples of modifiable risk factors including smoking, alcohol consumption and obesity. These represent targets for primary prevention strategies which can significantly reduce the risk of an individual developing CVD.

2.1CVDRisk Factors

A non-systematic literature search was conducted for known CVDrisk factors. These are shown in the following table, with associated references (Table 1):

Table 1: CVDrisk factor list

2.1.1Age

Age is the most significant non-modifiable risk factor for developing cardiovascular disease. Data from the WHO and UN suggest that the risk of mortality from ischaemic heart disease triples with each decade of life (2.3-2.7-fold increase per decade for men, 2.9-3.7 for women).[3] Similar gradients are seen in carotid disease and peripheral arterial disease,[4-8] with risk of developing these conditions increasing exponentially with age. The mechanisms for this are debatable, but may includediminishing structural and mechanical integrity of the heart and blood vessels as well asthe increased risk of exposure to other common risk factors.

2.1.2Gender

Men are at higher risk of developing CVD than women, although the risk for women increases substantially after menopause.[9-20]Over time and at different ages, independent of diagnostic and treatment practices, women have a similar or slightly higher prevalence of angina than men across countries with widely differing myocardial infarction mortality rates.[19]There are also gender-specific differences in the effects of other risk factors.Guideline-recommended treatments for angina are underused in women and older patients.[21] These suboptimal practice patterns, which are worst in older women, are of particular concern. The prevalence of minor ECG changes is slightly higher among men (10.4%v 9.5% in women). The occurrence of ischaemia-like findings on the ECG was comparable between men and women (9.0% v 9.8%).[20]

2.1.3Ethnicity

Migrants of South Asian descent worldwide have elevated risks of morbid and mortal events because of ischaemic heart disease.[2223]In the UK, mortality from IHD in both South Asian men and women is 1.5 times that of the general population, and South Asians have not benefited to the same extent from the general decline in deaths caused by IHD over the last few decades.[2425] Declines in stroke incidence have been observed the UK in men, women, White groups, and those aged >45 years, but not in Black groups. The reduction in prevalence of before-stroke risk factors was mostly seen in White patients aged >55 years, whereas an increase in diabetes mellitus was observed in younger Black patients.[2627]Incidence rates of first ever stroke adjusted for age and sex are twice as high in Black people compared with White people. This excess incidence cannot be accounted for by differences in social class in ages 35-64. Black people tend to have their first stroke at a younger age than White people. The excess incidence is found in all pathological types of stroke but is greatest for primary intracerebral haemorrhage.[28]

2.1.4Hypertension

Hypertension ranks among the most significant risk factors for cardiovascular disease. Willey et al. estimated hypertension to account for almost one quarter of the population risk of CVD (population attributable risk (PAR) = 24.3%, 95% CI 13.2-35.4%).[29] Severe hypertension (systolic blood pressure, SBP≥160mm Hg) has been associated with a two-fold increase in risk of CHD.[30] The risk among pre-hypertensive individuals (BP 120-139/80-89 mm Hg) is debatable, with different studies drawing different conclusions. However, Huang et al. recently performed a meta-analysis including over 500,000 individuals from 17 studies and demonstrated a significant association of pre-hypertension with risk of CHD (RR 1.43, 95% CI 1.26-1.63).[25]Hypertension has a stronger association with stroke, with PAR estimates as high as 36%.[5]Sun et al. showed an increased risk of stroke across several different categories of hypertension, with combined systolic and diastolic hypertension presenting the greatest risk (OR 2.13, 95% CI 1.78-2.75).Similarly for PAD, Fowkes et al. calculated an odds ratio of 1.47 (95% CI 1.37-1.57) among hypertensive patients (BP ≥ 140/90 mm Hg) compared to normotensive subjects.[7]

2.1.5Smoking

Smoking has long been established as a risk factor for CHD, and represents perhaps the most easily modifiable risk factor for CVD. In ranking the major risk factors for coronary heart disease, Schnohr et al. determined that smoking was the most important risk factor for women (OR 1.74, 95% CI 1.43-2.11) and the fifth most important for men (OR 1.62, 95% CI 1.25-2.09) when comparing non-smokers to individuals smoking >15g of tobacco per day.[30] Similarly, Tolstrup et al. showed a significant correlation between the amount of tobacco smoked and the odds of developing coronary heart disease in a pooled analysis, and that the risk was higher for women than in men.[31] Furthermore, there is ample evidence suggesting that cessation of smoking greatly reduces the risk of developing CHD.[32] Similar associations are seen in stroke, with Peters et al. reporting an 83% increased risk of stroke among women who currently smoke compared to those who have never smoked, and 67% increase among men.[13] There is an approximate two-fold increase in the risk of PAD for both current smokers (OR 2.09, 95% CI 1.91-2.29) and former smokers (OR 1.87, 95% CI 1.64-2.18) compared to non-smokers.[7]

2.1.6Diabetes

Diabetes is associated with a 2 to 4-fold increase in the risk of developing cardiovascular disease.[78112933-35] While this association may be somewhat confounded by the fact that diabetes is linked to several other CVD risk factors (e.g. obesity, hypertension, dyslipidaemia), there is evidence to suggest that diabetes is an independent risk factor. There is a graded positive association between HbA1c levels and risk of CHD, which holds true after adjustment for blood pressure, BMI and blood lipids, suggesting that poor blood glucose control increases risk of CHD.[36] (Zhao et al. 2014). Yusuf et al. estimated the odds of an individual with diabetes having a heart attack are 2.37 times greater (99% CI 2.07-2.71) than a non-diabetic when adjusting for other factors.[34]

The risk for stroke is similarly elevated, as studies show an approximately two-fold increase in the risk of stroke in diabetics compared to non-diabetics. Data from the Clinical Practice Research Datalink (CPRD) in the UK and estimated an odds ratio 2.19 (95% CI 2.09-2.32).[37] Peters et al. determined a similar risk in a recent meta-analysis, while noting an increased risk in women compared to men (OR 2.28, 95% CI 1.93-2.69 vs. 1.83, 95% CI 1.60-2.08).[12] Diabetes is also a key risk factor for peripheral arterial disease. Estimates range from 10-40% for the prevalence of PAD among individuals with diabetes, and PAD is the leading cause of amputation in diabetics.[38] Fowkes et al. showed a 68% increased risk of PAD among diabetic subjects in their meta-analysis (OR 1.68, 95% CI 1.53-1.84).[7]

2.1.7Obesity

We are now beginning to understand the underlying mechanisms as well as the ways in which smoking and dyslipidaemia increase, and physical activity attenuates, the adverse effects of obesity on cardiovascular health.Obesity is associated with other risk factors – diabetes, dyslipidaemia etc- which may mediate its effect on CVD. [39]Adipose tissue releases a large number of bioactive mediators that influence not only body weight homeostasis but also insulin resistance - the core feature of type 2 diabetes - as well as alterations in lipids, blood pressure, coagulation, fibrinolysis and inflammation, leading to endothelial dysfunction and atherosclerosis.[40]Body mass index (BMI) shows a positive association with cerebrovascular risk which is non-significant after adjustment for physical inactivity, smoking, hypertension, and diabetes (odds ratio 1.18; 95% CI, 0.77 to 1.79).[41] Markers of abdominal adiposity are strongly associated with the risk of stroke/TIA. For the waist-to-hip ratio, adjusted odds ratios for every successive tertile were greater than that of the previous one.

2.1.8Dyslipidaemia

High total cholesterol, low high density lipoproteins (HDL), and high low density lipoproteins (LDL) are all well-established CVD risk factors and are included in both the Framingham and QRISK2 CVD risk scores.[42-44]

2.1.9Deprivation and socioeconomic status

In the UK material deprivation is a well-established CVD risk factor and is included in the QRISK2 CVD risk score.[4546] Favourable population-wide trends in smoking, blood pressure and cholesterol are consistent with falling CHD death rates. However, adverse trends in obesity and diabetes in deprived populations are likely to counteract some of these gains. Furthermore, little progress over the last 15 years has been made towards reducing risk factor inequalities.[47] Approximately half the recent CHD mortality fall in England is attributable to improved treatment uptake. This benefit occurred evenly across all social groups. However, the opposing trends in major risk factors means that their net contribution amounted to just over a third of the CHD deaths averted; and these also varied substantially by socioeconomic group.[48]

2.1.10Physical Activity

In a recent meta-analysis, compared with individuals reporting no leisure time physical activity, there was a 20% lower overall mortality risk among those performing more than the recommended minimum, a 31% lower risk at 1 to 2 times the recommended minimum, and a 37% lower risk at 2 to 3 times the minimum. A similar dose-response relationship was observed for mortality due to CVD.[49] Sedentary time is independent of physical activity associated with an increased risk of diabetes, CVD and CVD mortality.[50]Compared with inactive individuals, those who exercise for an average of 15 min a day have a 14% reduced risk of all-cause mortality, and a 3 year longer life expectancy.[51].

2.1.11Inflammatory Markers

Although there is a growing evidence base showing the importance of other inflammatory marker data,[52]they are either not available in the national data sources available for use, or (more commonly) data are not available for small local populations, so we have not considered them further here.

2.1.12Chronic Kidney Disease

Chronic kidney disease (CKD) is associated with a number of other CVD risk factors, notably hypertension and diabetes, which contribute to the increased risk of CVD observed in CKD patients. However CKD has been shown to be an independent risk factor for CVD after adjustment for these related factors.[53]There is a relative risk of 1.4 (95% CI 1.3-1.5) for CHD development in individuals with CKD.[54]The data on stroke is somewhat less clear, with different studies arriving at dissimilar conclusions on the independence of CKD as a risk factor.[55] However in a large meta-analysis, Lee et al. showed a statistically significant 43% increase in risk of stroke among patients with a glomerular filtration rate (eGFR) < 60 mL/min/1.73m2 (RR 1.43, 95% CI 1.31-1.57), which held even after adjustment for traditional risk factors.[56]CKD is an established risk factor for PAD, with a two-fold increase in odds of developing PAD among patients with an eGFR < 60 mL/min/1.73m2 (OR 2.0, 95% CI 1.4-2.7).[6]

Table 2 summarises all the CHD risk factors we reviewed with their recent pooled, matched or adjusted odds ratios.

Table 2: CVD risk factors with their pooled, matched or adjusted odds ratios

Peters et al have conducted several systematic reviews, one for each major stroke risk factor.[1213] Table 3provides a summary of stroke risk factors specifically with their pooled, matched or adjusted odds ratios from these and other reviews.

Table 3: stroke risk factors with their pooled, matched or adjusted odds ratios

2.1Stroke prevalence from the literature

Worldwide, a systematic analysis for the Global Burden of Disease (GBD) Study 2013 showed CVD made up 21,177 (14 947 to 28 436) years of life disabled in 2013.[1] In the UK, GBD showed that the main drivers for improvement in life expectancy in nearly all countries and in the English regions have been declines in CVD and, to a lesser extent, cancer mortality.[61]Throughout the UK, prevalence of stroke in men was almost twice as high than that for women in 2013, although this data was obtained from GP diagnoses in the Clinical Practice Research Datalink (CPRD), which underestimates actual prevalence (Table 4).[2] Applying country-specific and age-specific population estimates, obtained from national statistics agencies, to prevalence data from CPRD suggests that >915 000 people in the UK have suffered an MI and >1.3 million are living with angina.[2] Consequently, if we combine estimates for MI and angina, we find that almost 2.3 million people in the UK are living with some form of CHD.[19]

Table 4: prevalence of cardiovascular conditions, UK 2013 (from CPRD)

Prevalence estimates of stroke worldwide are shown in Table 5.